ASA 128th Meeting - Austin, Texas - 1994 Nov 28 .. Dec 02

2aAO3. Inversion of acoustic field data using genetic algorithms: Shallow-water results.

D. F. Gingras

P. Gerstoft

SACLANT Undersea Res. Ctr., Viale San Bartolomeo, 400, 19038 La Spezia, Italy

Precise knowledge about environmental parameters is highly desirable for precise acoustic modeling of the underwater channel. Simulations have shown that global inversion methods such as simulated annealing and genetic algorithms are effective for the estimation of both geoacoustic and geometric (source location, receiver locations, and water depth) parameters. In this paper genetic algorithms are applied to acoustic field observations. The field observations were collected at a shallow-water site in the Mediterranean Sea near the Italian coast, where the depth was approximately 130 m. The source was stationary and produced a broadband signal at 170 Hz; a vertical array of receivers which spanned most of the water column was employed. The inversion process used the Bartlett processor for the object function. A range-independent version of the SACLANTCEN normal-mode acoustic propagation model snap was used for forward modeling. Inversion was carried out simultaneously over source location parameters, range and depth, and environmental parameters. A 20-min sample of observations was considered; at 1-min intervals estimates of the parameters were computed. As a function of both time and frequency, over a limited frequency band, it was shown that the genetic algorithms produced stable estimates for both the environmental and source location parameters.